A lyft driver told me about this today, and played me his digital avatar. It was better than I expected.
"Create a digital voice that sounds like you with only one minute of audio. Simply sign up, record yourself for at least one minute and you will be able to generate any sentence you like with your own digital voice."
The 1 minute statement isn't accurate. They require 30 sentences, which takes about 5 minutes. The cleaner and more abundant your input, the better the output.
https://lyrebird.ai/vocal-avatar
That's pretty cool, but I am not about to knowingly give my voice over to faking software.
Along these lines Wired just put together a documentary on AI which features the tech for deep fakes. (https://www.youtube.com/watch?v=ZJixNvx9BAc)
Also related:
AI Can Now Detect Deepfakes by Looking for Weird Facial Movements (https://www.vice.com/en_us/article/evy8ee/ai-can-now-detect-deepfakes-by-looking-for-weird-facial-movements)